Research outline


Complex machines have been part of our landscapes for many centuries and have come to dominate land, sea and air since the Industrial Revolution. Edward O. Wilson described our current age of mass extinction as the ‘Age of Loneliness’ and in many ways our technologies in these shared environments have been technologies of loneliness, that do not serve life but industry. Terms such as biosphere and technosphere are used to describe the world, but outside the front-door these are hard to find. Out there its all hybrid. This may seem obvious, but human tools, infrastructures and machines that are present in the landscape still embody this redundant dichotomy. Even cars - our primary way of transversing complex landscapes - are designed by engineers, not by ecologists, but now they are starting to drive themselves. Will they teach temselves about the landscapes and the many different beings in that shared space?


Environmental Machine Learning

Until very recently the ability to relate to the environment was limited to plants and animals, but now machines are starting to blur those lines. Artificial Neural Networks can be ‘trained’ using vast data sets. Over time it recognises dogs, people, etc through machine learning. Environmental Machine Learning (EML) explores if / how this synthetic 'world-view’ relates to the ‘umwelt’ that biological creatures experience. In a program of guest speakers and fieldwork sessions Random Forests aims to explore experimental systems as vehicles for materialising questions. These will be addressed with critical reflection in print and web media, like a field guide to environmental machine learning.



Some first applications for machine learning are being developed by ecologists as a way to address big data and heterogeneity issues in their data. (Ecologists deal with anything from genetic data, to climate, or species abundance.) It is deployed to extract wildlife sightings from the web, identify species in pixels and soundfiles. EML aims to take a less task oriented view; what the appearance of machine learning in biodiverse environments could mean. What does it mean if machines join animals and plants there on more equal levels of awareness? Some artists, designers, environmentalists and conservationists have started probing those questions. EML aims to bring those people together to map the territory, draw the first outlines of environmental machine learning and dig out the more fundamental questions it raises.



The question how human technology relates to wild systems is central to the health and future development of society. All mayor tech companies have made AI their top priority, and this sets the scene for AI to be developed within the framing of company oriented goals. Much of the current critical reflection focusses on the impact of AI on human labour, human privacy and human war. EML asks what the impact of AI is beyond our species. If the most pressing issues for society relate to environmental processes, that is one motivation for this research: with ecosystems collapsing we need an age of harmonising technologies urgently. The more fundamental motivation is simply curiosity for this newly forming territory: exploring the interactions between animal, machine and environment. And to examine if/how their ways to learn through experience relate.


note: Random Forests are a type of analysis in machine learning in which a large number of simpler operations called 'Decision Trees' are combined. This 'Random Forest analysis' only uses portions of the selectors (at random) on the data, so individual trees can vary more, which increases accuracy.